Explore, Chat, Watch, Memory
The core feedback loop in OpenStar.
OpenStar is a loop, not a set of disconnected tabs.
Explore presents repo candidates shaped by public corpus data and OpenStar Memory.
Chat investigates high-intent questions with tools, repository evidence, and user context.
Watch tracks repositories that deserve return attention.
OpenStar Memory records taste and feedback signals that improve the next feed.
Useful actions include save, watch, hide, ask, open, and priority feedback. These actions should make future recommendations clearer instead of resetting the experience.
Explore should generate candidate repositories and explain why they may fit.
Chat should inspect a smaller set of repos deeply and keep tool evidence visible.
Watch should turn important repo changes into low-noise briefs.
Loop health
| Signal | Healthy behavior |
|---|---|
| Explore | User sees relevant repos without only seeing prior stars. |
| Chat | User can inspect evidence and move decisions into Save or Watch. |
| Watch | User gets useful return-value briefs instead of raw noise. |
| Memory | User can understand and tune why recommendations changed. |